Matchmaking Rating
A matchmaking rating is the number a game keeps for each player to estimate their skill, so its matchmaker can build balanced matches. A rating system decides how that number starts, how much it moves after each match and how certain it is, often tracking a confidence value beside the rating. Its design choices, such as how it handles teams, parties and new players, shape how fair matches feel.
Also called
MMR, matchmaking rating system, skill rating
,
What Does a Matchmaking Rating Track?
A single number can say how good a player seems. It can't say how sure the system is. Many rating systems in use today track both: an estimate of skill and a measure of confidence in it.
Glicko-2 stores a rating, a rating deviation and a volatility. A new player starts at 1500 with a deviation of 350; the deviation shrinks as results come in and grows again while a player is away (Glickman, March 2022).
TrueSkill, from Microsoft Research, stores a skill estimate (μ) and an uncertainty (σ), and displays a conservative μ minus 3σ so a player with few matches isn’t overrated (Microsoft Research).
The confidence value does much of the work. While it’s wide, each result moves the rating a lot, so a new player finds their level quickly. Once it narrows, the rating moves slowly, so one bad evening doesn’t undo months of play. Elo, the best-known system, tracks only the estimate, which is one reason many games extend it.
How Does a Matchmaking Rating Handle Teams and Parties?
In a team game, one result updates every player on the team, so each match says less about any one of them. Microsoft Research's TrueSkill table shows the gap: at least 12 matches to identify a player's skill in one-on-one games, and at least 91 in two teams of eight, with the real number up to three times higher (read 29 October 2026).
Two common responses, each with a trade-off:
Weigh individual performance. TrueSkill 2 adds individual scores to each player’s update. Ratings settle faster, but players can start playing for their stats rather than the win, so the stats chosen matter.
Keep separate ratings per mode or queue. Skill in solo play doesn’t always carry over to team play, or the reverse.
Parties raise their own question: how to rate a group that queues together. Averaging the members is common, though an average can hide a wide gap between the strongest and weakest player. How strictly to match parties is a skill-based matchmaking choice.
How Do New Players, Seasons and Inactivity Change a Rating?
Placement matches, season resets and rating decay look like three separate features. In systems that track confidence, they're often the same lever: widen the uncertainty, and the rating moves faster until it settles again.
New players start with wide uncertainty, so their first matches, often called placement matches, move them quickly toward their level.
Players returning from a break may have changed. Glicko-2 widens a player’s deviation for every rating period they miss, without changing the rating itself.
A new season can soft-reset everyone: pull ratings part of the way back toward the middle, or widen their uncertainty, so the ladder reshuffles without starting from zero.
Ratings also drift over time as players join and leave, which is another reason many games reset on a schedule. Players usually see these effects through a visible rank or tier rather than the rating itself, since many games keep the rating hidden.
Where Should a Matchmaking Rating's Results Come From?
In a match's lifecycle, the rating is read at the start and written at the end:
The game’s backend attaches each player’s rating to their matchmaker ticket.
The matchmaker groups tickets whose ratings fit its rules.
The match plays on a server.
The result is reported, and the backend updates each rating.
Step 4 is where fairness can break. A result reported by players’ clients can be faked, so many games take it from the authoritative server that ran the match. Leavers, boosting (a stronger player raising someone else’s account) and smurfing (an experienced player on a new account) are the edge cases worth deciding early.
On Edgegap, the rating stays in the studio’s backend. It reaches the matchmaker as a ticket attribute, sent server to server when it’s sensitive, and a number-difference rule matches on it, for example within 50 points, widening to 150 after 30 seconds (docs). Edgegap’s matchmaker has SDKs for Unity and Unreal Engine.
Players of skill-based games often expect matchmaking on par with the largest studios’, whatever the size of the team behind the game. Researching the rating system a game needs before launch pays off; the Call of Duty deep dive shows how one large studio weighs its trade-offs.
A word from our sponsor (ourselves!)
A match isn't fair when one player lives next to the datacenter. Edgegap's Edge Cloud draws on up to 615+ distributed locations on demand, placing each server based on where every player in the match connects from rather than in a fixed region. In Edgegap's 1v1 study, that narrowed the fairness gap between opponents by 28%.
